Better data artifact management

[superdosh]

Want some sort of data runway that enables:

• All the various input types (dvc, local file, in-memory dataframe, mlflow artifact)

• Handles mlflow artifact logging (always write to mlflow)

• Should be an argument to other runways (but leave the easy path)

• Should expose properties so someone can easily share artifacts

• Include an "empty" flightpath -- data input + logged output